MMRReranker: Maximal Marginal Relevance (MMR) Reranker

In VectrixDB: Lightweight Vector Database with Embedded Machine Learning Models

Maximal Marginal Relevance (MMR) Reranker

Description

Reranks for diversity using MMR algorithm

Public fields

lambda

Balance between relevance and diversity (0-1)

Methods

Public methods

• MMRReranker$new()

• MMRReranker$rerank()

• MMRReranker$clone()

---

Method new()

Create a new MMRReranker

Usage

MMRReranker$new(lambda = 0.7)

Arguments

lambda

Relevance vs diversity tradeoff (higher = more relevance)

---

Method rerank()

Rerank for diversity

Usage

MMRReranker$rerank(query_vector, doc_vectors, doc_ids, scores, limit = 10)

Arguments

query_vector

Query embedding

doc_vectors

Matrix of document embeddings

doc_ids

Vector of document IDs

scores

Original relevance scores

limit

Number of results

Returns

Data frame with reranked results

---

Method clone()

The objects of this class are cloneable with this method.

Usage

MMRReranker$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

VectrixDB documentation built on Feb. 20, 2026, 5:09 p.m.